Poverty and Plant-Based Diets

Forty years ago, Frances Moore Lappe wrote Diet for a Small Planet, a combination cookbook and food industry critique. In it, she pointed out that the grain we feed to livestock animals could instead be fed to hungry people.

The recent shock in food prices has led to increased examination of food cost determinants, and the data provides interesting insights into how our diets can affect the lives of the world's poor.

The Numbers

According to Counting Animals, a vegetarian saves 29 chickens, 1/2 of a pig and an eighth of a cow each year. Using the formula developed by Fortenbery and Park, 9 million such vegetarians would reduce the price of corn by $5/bushel. Using this as a proxy for soy, food prices of the ten staple foods would drop by 20%1. This corresponds2 to the central scenario of Dessus et al.,estimated to cause 233.2 million people to come out of absolute poverty (defined as living on less than $2/day). Using Goklany's estimates, this would avert 1.22 million deaths, and 42.7 million disability adjusted life-years.3

To put it in personal terms: one vegetarian saves one human for every eight years they're veg, and averts four DALYs per year of vegetarianism.

Cost Effectiveness

EAA has previously estimated that the top charities create one vegetarian-year for around $11. This means that top veg charities save one person for $90, and spend around $2.75 to avert a DALY. For comparison, the Against Malaria Foundation, GiveWell's current top pick, spends $2,300 per life saved or between $29 and $169/DALY.

Even with the generous padding that these rough calculations deserve, veg charities may be competitive with other poverty-focused charities.


Code used to calculate these numbers can be found here.
  1. This would cause a drop in soy and corn prices of 63%. However, these foods make up only a third of total global staples, meaning that aggregate staple price would drop by only ~20% (ceteris paribus). Note that Fortenbery and Park's model probably wouldn't handle such a large change well, so this should be considered a very rough estimate.
  2. Dessus and Goklany both examined the other direction: how many more people would enter poverty as the result of increased food prices. I assume here that the change is symmetric, i.e. the badness caused by an increase of $x is the same as the goodness caused by a decrease of $y
  3. Goklany separates DALYs meaning "disability with no death" from actual deaths, in contrast to places like GiveWell, which usually include premature death in their DALY calculation.


  1. "9 million such vegetarians would reduce the price of corn by $5/bushel"

    That >1% of heavy meat eaters giving up meat would drop the price of corn from $6.40 to $1.40 sounds like 10x or even 100x more of a decrease than we should expect. "Fortenbery and Park's model probably wouldn't handle such a large change well, so this should be considered a very rough estimate" isn't anywhere near a big enough disclaimer.

    1. Hey Jeff,

      Thank you for your comment. I apologize for not updating this blog - the model has gone through several iterations. You can look at an R script here which will allow you to calculate the effectiveness for an arbitrarily small number of new vegetarians, for which the F&P model better predict: https://gist.github.com/Xodarap/4699242

      A document discussing the methodology can be found here: https://docs.google.com/document/d/1TTd2HV6PhfDe8ZbN4wzndo1gU6TItqkV-bNGtJGve6k

      I think you are right to point out that the F&P regression is locally accurate but price is non-linear; I look forward to your comments about the improved model.

    2. It sounds like what you're doing is:

      1. Estimate the cost to convince someone to be vegetarian. You're using Brian Tomasik's $11/year which I think is too low, but let's ignore that for now.

      2. Estimate how much a person stopping eating meat decreases their indirect corn consumption. It sounds like you get 0.59 million bushels per year per 0.1M vegetarians, or 6 bushels per vegetarian. (Cell A5). This seems much too low.

      3. Estimate the effect on the price of corn based on a change in demand. You're reading F&P to say that if demand for corn changes by 1M bushels than the price of corn will change by $0.868. The US corn crop is around 10B bushels/year, so this is another way of saying that if corn demand changes by 0.01% then it's price will change by $0.868. This seems far more of an impact than we should expect. I'm not sure that F&P's 0.868 means what you think it does.

      4. Estimate how many people benefit from the drop in corn prices and how much they benefit. Let's ignore this for now too.

  2. Fascinating, but would love to see this commented on and verified further.

    1. I'd also be interested in:

      1. If there are shifts in choices of meats eaten that can have partial or similar effects.

      2. If this is about adopting a vegetarian, or vegan diet. It seems like in this scenario, eating eggs is not much better than eating chicken.

    2. Thanks Josh. I've since learned about this interesting paper – it's in German but Google translate works pretty well. Basically they analyze how a drop in meat consumption in the OECD countries affects global food prices. (Table 5 on page 125 is the main thing to look at.)

      They find that meat supply is much less elastic to demand than I had been assuming, so I think the results in this blog post are overstated.

  3. Assuming that a model for short-term changes in corn prices developed for estimating the effects of ethanol production holds for long-term changes in feed consumption levels, and assuming that if an increase in food prices drives a certain amount of people into poverty then a decrease of the same size will pull just as many people out of poverty, and assuming that the deaths and DALYs prevented by pulling some people out of poverty are equal to those lost due to the same number of people descending into poverty... then we can take these numbers as a rough and generous estimate of the effects of vegetarian charities.

    The Fortenbery and Park's model found that changes in ethanol production drove almost all the activity of their equation, with feed consumption "statistically insignificant". Creating more vegetarians exerts influence on corn usage only through decreasing feed consumption.

    The Dessus paper finds that almost all the additional cost of increased poverty is due to deepening the poverty of the existing poor, not due to new poor created. This breaks the connection to Goklany's estimates, which are based on the number of deaths and DALYs lost per million in poverty. That is, Goklany assumes that a poor person is a poor person and all suffer equally, but Dessus just told us that the new poor created by a price increase are only very slightly poor, which shouldn't be expected to suffer as much as the average person in poverty.

    So the conversion from decrease in livestock to decrease in corn price is deeply flawed, the conversion from food prices to increased poverty levels is misleading, the conversion from increased poverty to increased suffering is deeply flawed, and the assumption that increases in poverty and suffering correspond to equal decreases in poverty and suffering is entirely without support. I feel like this renders the cost-per-DALY number meaningless, and no amount of generous padding of the final product can help. (Comparing it to GiveWell's predicted[1] cost-per-DALYs particularly egregious; if you allowed me to use your methodology on fighting malaria I could probably quote you cents-per-DALY or better.)

    1: the zoho doc is not public, but I presume it contains a conversion from lives saved to DALYs

    1. Thanks for the comments. I'm not sure I understand everything you're saying (e.g. the fact that were looking only at a subset of the affected people should cause us to underestimate the benefits, not overestimate) but see my above comment to Josh regarding new research which has come out since this post.